KHelper V3 User Guide for VTubers & Streamers

1. Operating System: Windows 10 64-bit (version 1909 or newer) or Windows 11 64-bit.
2. Processor (CPU): Quad-core processor or higher (such as Intel Core i5 or AMD Ryzen 5).
3. Memory (RAM): 8 GB minimum; 16 GB or more is recommended when running OBS Studio and Live2D / 3D tracking software at the same time.
4. Storage Space: Including the app and AI separation engine, the GPU version requires ~5 GB of free disk space and the CPU version requires ~1 to 2 GB (can be moved to another drive in Settings).
5. Graphics Card (GPU Hardware Acceleration):
   • Supports NVIDIA GeForce GPUs with 4 GB+ dedicated VRAM (RTX 50/40/30/20 and GTX series). With CUDA acceleration enabled, separating a song takes roughly 15 to 35 seconds.
   • On systems with AMD GPUs, Intel integrated graphics, or laptops without a dedicated GPU, KHelper runs in [Standard CPU Processing Mode] with all features available.

1. Go to the GitHub Releases page and download the latest Windows installer (KHelperV3-Setup-x.x.x.exe).
2. Double-click the installer. If Windows displays a blue "Windows protected your PC (SmartScreen)" prompt, click [More info] and then [Run anyway].
3. Follow the installer prompts and launch KHelper V3 from the desktop shortcut.
4. First-Time AI Separation Engine Setup:
   • The first time you separate a song, the AI Vocal Separation Engine Setup dialog opens and detects your hardware.
   • If you have a compatible NVIDIA GPU, click [Install & Enable GPU Acceleration] (downloads ~2.7 GB of PyTorch CUDA dependencies). Otherwise, click [Install CPU Engine] (~600 MB). Once installed, the runtime stays on your PC and does not need to be re-downloaded after app updates.
5. Moving Storage Locations (Optional):
   • By default, your song library and separation engine are stored on the C: drive (AppData). If C: drive space is tight:
     • To move your song library: Go to [Settings] → [Storage] and click [Change Storage Location / Migrate...] to choose a folder on another drive.
     • To move the separation engine: Go to [Settings] → [Vocal Separation], and under Separation Engine Storage Location, click [Change Storage Location...].

1. Use the top search bar (Search songs / artists) to search YouTube and download a song, or click [Library] in the left sidebar → [+ Add Song] in the top right to import local audio files.
2. Find a song marked as Original in your library and click [Separate] (or right-click the song and select [Start Separation]).
3. Wait until the audio status changes to Separated, then double-click the song to play it.
4. Put on your headphones (before changing audio settings, sound plays through your current Windows System Default playback device) and move the [Instrumental] and [Vocal] sliders on the bottom player bar to confirm you can control the backing track and guide vocal volumes independently. Next, proceed to Chapter 01 to set up dual-track routing for OBS.

During a karaoke stream, hearing a quiet guide vocal in your headphones helps you stay on beat and in tune, while your stream audience should only hear the instrumental backing track and your microphone.

KHelper V3 provides two separate audio output channels:

1. [Audience Output (Stream)]: Outputs only the separated instrumental backing track for OBS to broadcast to viewers.
2. [Headphone Output]: Outputs both the instrumental backing track and the guide vocal to your headphones.

If you do not have a multi-channel audio interface with independent loopback routing, you can install the free VB-CABLE virtual audio driver to connect KHelper to OBS:

1. Visit the official VB-Audio website (https://vb-audio.com/Cable/) and download VBCABLE_Driver_Pack.zip for Windows.
2. Extract the ZIP file, right-click VBCABLE_Setup_x64.exe, select [Run as administrator], and complete the installation.
3. After installation finishes, restart your Windows PC.
4. After rebooting, Windows Sound Settings will include a new virtual device pair:
   • Playback device: CABLE Input (VB-Audio Virtual Cable)
   • Recording device: CABLE Output (VB-Audio Virtual Cable)

Open KHelper and go to [Settings] → [Audio Output]:

1. [Audience Output (Stream)]: Select CABLE Input (VB-Audio Virtual Cable).
2. [Headphone Output]: Select your physical headphones or audio interface (Note: [Audience Output (Stream)] and [Headphone Output] must be set to different devices. If both use the same device, KHelper automatically mutes [Audience Output (Stream)] to prevent doubled audio).
3. Guide Vocal Volume: Adjusting the [Vocal] slider on the bottom player bar only changes the guide vocal volume in your headphones and never affects [Audience Output (Stream)].

1. In the OBS Sources dock, click [+] → [Audio Input Capture], and name it KHelper Backing.
2. Set Device to CABLE Output (VB-Audio Virtual Cable) and click [OK].
3. Play a song in KHelper and verify that the KHelper Backing meter moves in the OBS Audio Mixer.
4. ⚠️ Important: Disable OBS Desktop Audio (Prevents Guide Vocal Bleed)
   • By default, OBS captures Windows Desktop Audio. If left enabled, the guide vocal playing in your headphones will be captured by Desktop Audio and broadcast to your viewers.
   • In OBS, go to [Settings] → [Audio] → Global Audio Devices, and set both Desktop Audio and Desktop Audio 2 to [Disabled].
   • Confirm that the OBS Audio Mixer only shows your Microphone track and the KHelper Backing track.
5. Check OBS Advanced Audio Properties:
   • Click the gear icon in the OBS Audio Mixer and select [Advanced Audio Properties].
   • Make sure Audio Monitoring for both KHelper Backing and Microphone is set to [Monitor Off] so OBS does not loop audio back into your headphones.

Hearing your own voice in your headphones helps you control pitch and vocal volume:

1. External Audio Interfaces:
   • Enable the hardware Direct Monitor button on your interface to hear your microphone with zero latency.
2. Yamaha AG03 / AG06 Series Note:
   • Set the physical streaming switch on the mixer to Dry CH 1-2 or Stereo In. Do not set it to Loopback (Loopback mixes your headphone guide vocal back into your microphone input and sends it to OBS).
3. USB Microphones:
   • Plug your headphones directly into the 3.5mm headphone jack on the USB microphone and turn up the microphone's monitor/sidetone dial.
   • Avoid setting your microphone to Monitor and Output in OBS, as software monitoring introduces a noticeable delay that makes singing difficult.

1. Press Win + R on your keyboard, type mmsys.cpl, and press Enter to open the Windows Sound control panel.
2. Under the Playback tab, right-click CABLE Input and your headphones → [Properties] → [Advanced]. Under the Recording tab, right-click CABLE Output and your microphone → [Properties] → [Advanced].
3. Set the default format on all devices (and in OBS Audio Settings) to 24-bit, 48000 Hz or 16-bit, 48000 Hz (you can also check the Hz badge in the upper-right corner of each selected device card in KHelper [Settings] → [Audio Output]).
4. Using the same sample rate across all devices helps prevent audio crackling or sample clock drift during long streams.

Before going live, take 30 seconds to verify your routing:

1. Importing Local Audio Files:
   • Open [Library], click [+ Add Song] in the top right, and under Local Files click [+ Choose Files] (or drag and drop MP3, WAV, FLAC, or M4A files directly into the song list).
2. YouTube Search & Batch Link Import:
   • Use the top search bar to find songs on YouTube. Under YouTube Stream Results, you can preview tracks or click the download button to save them to your local library.
   • You can also open [+ Add Song], paste multiple URLs into YouTube links (one per line), and click [Check and Import] to queue them together.
   • Note: Online streaming tracks must be downloaded to your local library before pitch shifting, speed control, and vocal separation can be used.
3. Track Types (Original vs. Instrumental):
   • When importing or downloading a song, you can set its type to Original or Instrumental.
   • Songs marked as Original can be processed with AI vocal separation. If your audio file is already an off-vocal backing track, set its type to Instrumental to play it directly without separation.

You can set your default separation quality in [Settings] → [Vocal Separation], or right-click a song and select [Start Separation] to choose a mode:

1. [Fast]:
   • Uses fewer system resources and finishes fastest, with slightly lower separation quality. Suitable for quick previews. Uses ~25 MB of disk space per song.
   • NVIDIA GPU: ~15 seconds for a 4-minute song.
   • CPU Mode: ~2 to 3 minutes for a 4-minute song.
2. [Normal] (Recommended):
   • Balances processing speed and audio quality; recommended for standard karaoke streams. Uses ~35 MB of disk space per song.
   • NVIDIA GPU: ~25 to 35 seconds for a 4-minute song.
   • CPU Mode: ~4 to 5 minutes for a 4-minute song.
3. [High Quality]:
   • Takes longer to process, but provides the cleanest separation result. Uses ~50 MB of disk space per song.
   • NVIDIA GPU: ~45 to 60 seconds for a 4-minute song.
   • ⚠️ CPU Mode Warning: Running [High Quality] in CPU mode takes roughly 15 to 20 minutes per song under heavy CPU load. Do not run High Quality CPU separation while live streaming, as it can cause OBS frame drops.

1. Pre-Separating Your Setlist Before Streaming (Recommended):
   • Before going live, click [Separate] on each song in your planned setlist. Selected songs are added to the [Active Tasks] queue in the top right and processed one by one in the background.
   • Once separated, the vocal and instrumental stems are saved on your local drive. Playing a separated song during your stream simply plays the local audio files without running AI separation again.
2. Handling Live Song Requests on Stream:
   • With an NVIDIA GPU: Downloading and separating a requested song in [Normal] mode takes around 30 seconds while you chat with viewers.
   • With CPU-Only Processing: Avoid running AI separation while live on stream. Instead, take requests from your already-separated library or import pre-made instrumental tracks.

All AI vocal separation in KHelper V3 runs locally on your Windows PC. Your audio files are never uploaded to external cloud servers, and there are no usage quotas or subscription limits.

Every song spans a different pitch range. Singing notes above your comfortable limit causes vocal strain or voice cracks, while dropping a song too many semitones can make the verse too low to project clearly. By saving your comfortable vocal range, KHelper compares each song's vocal stem against your profile and suggests an appropriate key shift.

Go to [Settings] → [Vocal Range] and click [Set Up Range] (or [Modify Range]) to open Vocal Range Profile Setup:

1. Finding Your Range with the Interactive Piano (C2 to C6):
   • Clicking any piano key plays a reference tone in your headphones only (never sent to the stream output).
   • Click the [Lowest Note] card first, then click keys moving left from Middle C (C4) while singing along on an "Ah" vowel to find the lowest note you can sing comfortably without forcing your throat down (for example, G3).
   • Next, click the [Highest Note] card and move right from C4 to find the highest note you can sing steadily without straining (for example, D5).
   • Click [Save] in the bottom right.
2. Using Voice Type Presets:
   • You can also click one of the preset buttons at the top as a starting point and adjust from there:
     • [Soprano (C4~A5)]
     • [Mezzo-Soprano (A3~F5)]
     • [Tenor (C3~A4)]
     • [Baritone (G2~E4)]

1. After a song finishes AI vocal separation, KHelper automatically analyzes the vocal track in the background to detect its Lowest note and Highest note.
2. In Stream Mode, switching songs, opening the [Pitch] popup on the player bar, or clicking the microphone icon ([Show vocal range & recommended key]) in the top right of the lyrics box displays the Song Vocal Range pill.
3. Seeking to High and Low Notes:
   • Click the Lowest note or Highest note badge inside the pill to jump playback directly to the timestamp where that note occurs so you can preview it.
4. Applying the Recommended Key:
   • The right side of the pill displays a key recommendation button—for example, [⚡ Suggest -2] (shift down by 2 semitones) or [✨ Original Key (±0)] if the song already fits your range. Click the button to apply the pitch shift immediately.
5. Octave-Shift Recommendations for Wide or Distant Keys:
   • When a song sits far outside your range and is better sung an octave lower or higher, the button displays [⚡ Suggest +X (Sing 1 octave lower)] or [⚡ Suggest -X (Sing 1 octave higher)] (where +X / -X is the semitone transpose applied to the track, such as +3 semitones while you sing one octave lower), placing the climax notes within your comfortable range.
6. Dual-Track Pitch Sync & Estimation Note:
   • Pitch changes apply to both [Audience Output (Stream)] and [Headphone Output] simultaneously so the backing track and guide vocal stay in the same key.
   • Note: Song range is estimated automatically from the separated vocal stem (For reference only) and may be affected by backing harmonies, reverb, or separation artifacts. Click the high/low notes in the pill to seek and verify.

1. Recommended Range (±1 to ±3 Semitones):
   • Pitch shifts within ±3 semitones preserve natural instrument tone and vocal timbre.
   • Shifting by ±4 semitones or more introduces noticeable time-stretch coloration, so keeping adjustments within ±3 semitones is recommended when possible.
2. Adjusting Key While Singing:
   • You can click [Pitch] on the bottom player bar or in the [Mini Player] at any time during playback to adjust semitones using the - / + buttons or slider.

Many official music videos include 20 to 40 seconds of spoken dialogue or ambient sound before the song begins, creating dead air on stream. Cutting the audio file in an external editor normally breaks all timestamps in synced LRC lyrics. KHelper's built-in Audio Trim sets non-destructive start and end points and automatically shifts synced lyric timestamps to match.

1. Right-click a song in your library (or click the [More] ⋯ icon) and select [Edit Song Details].
2. Look at the Audio Trim waveform section at the bottom of the dialog (audio previews inside this window play in your headphones only):
   • Set Intro Start (In-point): Drag the left start handle on the waveform to where the music begins (lyric timestamp markers are shown along the waveform for reference), or click [Grab Current] while previewing.
   • Set Outro End (Out-point): Drag the right end handle to trim prolonged silence or outros at the end of the track.
3. Preview Your Trim Points:
   • Click [Play from Intro Start] to check the entry point, and click [Play 3.5s before Outro End] to check the ending.
4. Save or Reset:
   • Click [Save] to apply. Trimming only stores start and end markers in KHelper without modifying your original audio file on disk. You can click [Reset to Full] at any time to restore the full track length.

1. When you change Intro Start (In-point) and click [Save] on a song with Synced lyrics, KHelper automatically shifts all lyric timestamps earlier by the trimmed intro amount so subtitles stay in sync.
2. If any synced lyric lines fall inside the trimmed intro section, KHelper displays a Lyrics Conflict Detected prompt where you can choose:
   • [Drop from Synced Timeline (Recommended)]: Removes timestamps from the cut intro lines while preserving the raw lyrics text.
   • [Group at 00:00.00]: Places those intro lines at 00:00.00.

Tracks from different eras and sources (older anime CDs, YouTube uploads, modern streaming releases) are mastered at very different loudness levels. Without normalization, streamers have to constantly ride the OBS volume slider to prevent one song from being too quiet and the next from blasting viewers.

Go to KHelper [Settings] → [Audio Output], enable Volume Normalization, and choose a Target Loudness Preset:

1. [Standard (-14 LUFS · YouTube / Spotify Recommended)]: Suitable for general music playback or background BGM.
2. [Voice Priority (-16 LUFS · Twitch Mic Focus)] (Recommended for Karaoke Streams): Keeps backing tracks around -16 LUFS, leaving clear headroom for your live vocals.
3. [Gentle (-18 LUFS · Long Sessions / Fatigue Protection)]: Useful for long singing streams or softer vocal styles.
4. [Custom]: Lets you set a custom target between -20.0 and -10.0 LUFS.

When enabled, KHelper automatically standardizes perceived loudness across all songs and applies a built-in safety limiter to prevent clipping.

To avoid chasing volume sliders in two programs at once, set your baseline once in OBS and control live volume inside KHelper:

1. Layer 1: OBS Mixer Sets the Fixed Baseline (Set Once & Leave Alone)
   • During initial setup before going live, play a song in KHelper and watch the KHelper Backing meter in the OBS Audio Mixer. Set the OBS KHelper Backing fader to a comfortable baseline (for example, where the meter mainly sits within -20 to -14 dBFS, from the top of the green zone into the lower yellow zone).
   • Once set, leave the OBS fader untouched during your stream.
2. Layer 2: KHelper Controls Song-to-Song Balance & Live Volume Adjustments
   • Enable Volume Normalization in KHelper (recommended: [Voice Priority (-16 LUFS · Twitch Mic Focus)]) so songs from different sources start at a consistent loudness.
   • Whenever you want to turn the backing track or headphone guide vocal up or down before or during a song, adjust the [Instrumental] and [Vocal] volume sliders directly inside KHelper (on the bottom player bar or Mini Player) rather than changing faders in OBS.

On the OBS Audio Mixer meter (Green is below -20 dBFS, Yellow is -20 to -9 dBFS, Red is -9 to 0 dBFS), aim for these level ranges:

1. KHelper Backing Track:
   • Keep between -20 and -14 dBFS (top of green into lower yellow, averaging around -17 dBFS), with loud drum peaks staying below -12 dBFS.
2. Microphone Track (Speaking & Verse Singing):
   • Aim for -15 to -11 dBFS (middle-to-upper yellow zone, averaging around -13 dBFS), sitting roughly 4 dB above the backing track so lyrics remain clear.
3. Microphone Track (Chorus & High Notes):
   • Peaks reach -9 to -6 dBFS (lower red zone, roughly 8 to 10 dB above the backing track).
   • 💡 Helpful Note: OBS colors the meter red above -9 dBFS as a headroom warning—entering the lower red zone does not mean your audio is clipping (actual digital clipping only occurs when audio hits 0 dBFS). Hitting -9 to -6 dBFS during a loud chorus is normal.
4. Microphone Safety Limiter:
   • Right-click your Microphone track in OBS → [Filters], add a [Limiter] filter, and set its Threshold to -2.0 dBFS to prevent sudden loud peaks from clipping at 0 dBFS.

Under [Settings] → [Overlays], KHelper provides two types of browser overlays for OBS:

1. Lyrics Overlay: Displays real-time scrolling lyrics and pronunciation guides with customizable fonts, colors, stroke, glow, and animations (Scroll, Fade, Slide, Scale).
2. Setlist Overlay: Displays your live queue (Now Singing, Up Next, and Sung) with 12 built-in templates (Classic List, Record Card, Compact Strip, Neon Signboard, Countdown Board, Full List, Retro Pager, Cassette Player, Stage Marquee, Polaroid Setlist, Vertical Spacer List, and Spinning Disk List).

1. Copy the OBS Link:
   • In KHelper, go to [Settings] → [Overlays], and under Lyrics overlay designs click [Copy OBS Link] (or in Stream Mode, click the link icon in the top right of the lyrics panel → [Copy Lyrics OBS Link]). The default URL format is http://localhost:10001/obs/lyrics?design=default.
2. Add the Browser Source in OBS:
   • In OBS Sources, click [+] → [Browser].
   • Paste the copied link into the URL field.
   • Set the source Width and Height to the size you need.
   • Uncheck Close source when not visible (and Shutdown source when not visible) so lyrics do not lag or reload when switching OBS scenes.
   • Click [OK].

1. Copy the OBS Link:
   • In KHelper, go to [Settings] → [Overlays], click [Edit] under Setlist overlay designs to customize your template and click [Save], then click [Copy OBS Link] (or in Stream Mode, click the link icon → [Copy Setlist OBS Link]). The default URL format is http://localhost:10001/obs/setlist?design=default.
2. Add the Browser Source in OBS:
   • In OBS Sources, click [+] → [Browser].
   • Paste the setlist URL into the URL field, and set the source Width and Height to the size you need.
   • Uncheck Close source when not visible and click [OK].

KHelper can automatically generate phonetic reading guides above or below lyrics, including Japanese Furigana and Romaji, Chinese Pinyin, and Korean Romanization:

1. Quick Toggle in Stream Mode:
   • In Stream Mode, click the pronunciation toggle buttons at the top of the lyrics panel to turn Furigana and Romaji on or off for the current song.
2. Controlling Pronunciation Display on OBS Overlays:
   • Go to [Settings] → [Overlays] and click [Edit] on your lyrics design to open the Lyrics Overlay Editor.
   • Under Pronunciation Assist, set Ruby (top annotation) and Romanization (bottom line) to:
     • [Follow App]: Matches your toggle state in KHelper's Stream Mode.
     • [Show]: Always shows pronunciation guides on the OBS overlay.
     • [Hide]: Hides pronunciation guides on the OBS stream while letting you still view them inside the KHelper app.

Click [Lyrics Editor] in the left sidebar (or right-click any song and select [Edit Lyrics]):

1. Searching Online Lyrics:
   • Click [Search Lyrics] at the top, enter the song title or artist, click [Preview] to check the lyrics, and click [Use These Lyrics] to apply them (results marked Synced include time markers).
2. Importing Local Lyric Files:
   • Click [Import] to load a .lrc or .txt file from your computer.
   • Encoding Note: Save lyric files in UTF-8 encoding. If imported lyrics show garbled characters, open the file in Windows Notepad, select File → Save As, set Encoding to UTF-8, and import it again.
3. Aligning Plain Text Lyrics with Tap Mode (J):
   • If your lyrics are Plain text without timestamps, enable [Tap Mode] (Tap Mode (J)), play the song, and press the J key on your keyboard (or click [Tap]) as each line begins to stamp the current playback time. When finished, click [Save LRC].
4. Fixing Out-of-Sync Lyrics (Smart Align & Fine-Tuning):
   • Single-Line Adjustment: Use [-50ms], [+50ms], or [Apply Current Time] on individual lines in the editor.
   • Shifting or Scaling the Entire Song (Smart Align): If an imported LRC file is consistently early or late, click [Smart Align] at the top of the editor. Seek the player to the start of the first lyric line and click [Set Start (Offset)], then seek to the start of the last lyric line and click [Set End (Scale)]. KHelper automatically recalculates all lyric timestamps in between; click [Save LRC] when done.

For setup questions, bug reports, or feature requests not covered in this guide, reach out through our official channels:

• Official Discord Community: https://discord.gg/96zfTcBgZG
• GitHub Issues & Bug Reports: https://github.com/msfmsf777/karaoke-helper-v3/issues/new
• Official Website: https://khelper.live/